Estimating a social value set for EQ-5D-5L in Sweden

Abstract Background The study aims to elicit a value set based on the EQ-VT for the EQ-5D-5L that can be used to support decision-making in Sweden. Methods Participants were recruited from the general population based on age, sex and urban/rural area quota sampling from five regions across Sweden. In total, 785 interviews were conducted from February 2020 to April 2021 using the EQVT 2.1 protocol, and both composite time trade-off (c-TTO) and discrete choice experiments (DCE) were used to elicit health preferences. A variety of models have been tested for the c-TTO data (generalized least square, Tobit, heteroskedastic models) and DCE data (conditional logit model), as well as the combined c-TTO and DCE data (hybrid modelling). Model selection was based on theoretical considerations, logical consistency of the parameter estimates, and significance of the parameters (p = 0.05). Model goodness-of-fit was assessed by AIC and BIC, and prediction accuracy was assessed in terms of mean absolute error. The predictions for the EQ-5D-5L health states between models were compared using scatterplots. Results The preferred model for generating the value set was the heteroskedastic model based on the c-TTO data, with the health utilities ranging from -0.31 for the worst (55,555) to 1 for the best (11111) EQ-5D-5L states. Conclusion This is the first c-TTO-based social value set for the EQ-5D-5L in Sweden. It can be used to support the health utility estimation in economic evaluations for reimbursement decision making in Sweden.